Metadata-Version: 2.4
Name: thesean
Version: 0.1.6
Summary: Regression investigation harness for ML pipelines
Author: Danny Nguyen
License-Expression: MIT
Project-URL: Homepage, https://github.com/theseanai/thesean
Project-URL: Issues, https://github.com/theseanai/thesean/issues
Keywords: regression,investigation,ml,machine-learning,debugging,pipeline
Classifier: Development Status :: 4 - Beta
Classifier: Programming Language :: Python :: 3
Classifier: Programming Language :: Python :: 3.10
Classifier: Programming Language :: Python :: 3.11
Classifier: Programming Language :: Python :: 3.12
Requires-Python: >=3.10
Description-Content-Type: text/markdown
License-File: LICENSE
Requires-Dist: typer<1,>=0.9.0
Requires-Dist: pydantic<3,>=2.0.0
Requires-Dist: pydantic-settings>=2.0.0
Requires-Dist: jinja2>=3.1.0
Requires-Dist: tomli-w>=1.0.0
Requires-Dist: InquirerPy>=0.3.4
Requires-Dist: rich>=13.0.0
Requires-Dist: torch<3,>=2.0.0
Requires-Dist: numpy<3,>=1.24.0
Requires-Dist: scipy>=1.11.0
Requires-Dist: statsmodels>=0.14.0
Requires-Dist: textual>=0.80.0
Requires-Dist: tomli>=2.0.0; python_version < "3.11"
Provides-Extra: dev
Requires-Dist: pytest>=7.0; extra == "dev"
Requires-Dist: pytest-textual-snapshot; extra == "dev"
Requires-Dist: hypothesis; extra == "dev"
Requires-Dist: pytest-asyncio>=0.23.0; extra == "dev"
Requires-Dist: ruff>=0.4.0; extra == "dev"
Requires-Dist: mypy>=1.8.0; extra == "dev"
Dynamic: license-file

 # Thesean

 A regression investigation harness for ML pipelines, latent action systems, and world models.

 Thesean runs paired A/B evaluations across ML/world model experiments, detects regressions, and provides an interactive investigation workbench for debugging episode-level divergences.

 ## What it does

 - Configure A/B experiment pairs with different weights, planners, or configs
 - Run paired evaluations with live telemetry
 - Detect regressions via bootstrap significance testing across 5 metrics
 - Drill into individual episodes with signal timelines and event detection

 ## Quickstart

 Python 3.10+ required.

 ### Install

```bash
pip install thesean
```

 ### Prerequisites: Thesean requires an ML repo with a compatible adapter.

 The only available adapter is for [f1worldmodel](https://github.com/justinsiek/f1worldmodel).

 Clone the repo and follow its README to install dependencies.

 ### Run

cd into the ML repo project root:

Example:
```bash
  cd f1worldmodel
  thesean
```

 This launches the TUI. From there:

 1. Select or create a case.
 2. Pick a track, configure Run A (baseline) and Run B (candidate) with different checkpoints or planner settings
 3. Run the evaluation
 4. View the verdict and drill into episodes to investigate divergences

 ## Available Adapters

 | Adapter | Repo | Status |
 |---------|------|--------|
 | `f1` | [justinsiek/f1worldmodel](https://github.com/justinsiek/f1worldmodel) | Beta |

 ## Status

 Beta. APIs may change.

 ## License

 MIT
